RTI and Medusa Film vs Perplexity AI: the first Italian lawsuit for AI training

On December 3rd, 2025, RTI SpA and Medusa Film SpA (the broadcaster and the movie production company of Mediaset Group) filed a civil lawsuit before the Court of Rome against Perplexity AI Inc. for unlawful use of copyrighted content in the context of the AI training of Perplexity’s large language model (LLM), Sonar. 

The case between RTI/Medusa and Perplexity is based on the plaintiffs' allegation of a framework of multiple, systematic, and serious infringements of intellectual property rights, in violation of the core principles of European and national copyright laws that are applicable also in the generative AI ecosystem. The plaintiffs hold primary copyrights and related rights under Italian copyright law, as well as the trademarks that distinguish RTI’s programs and Medusa’s audiovisual works; such rights relate not only to content protected in its entirety but also parts, characteristic elements, frames, formats, and distinctive signs, all covered by exclusive protection.

RTI and Medusa highlight that they have full ownership of the exploited audiovisual content: works of authorship protected under Articles 1 and 2 of the Italian Copyright Law, as well as related rights of the producer and broadcaster under Articles 45, 78-ter, and 79 of the Italian Copyright Law (LDA). These rights expressly include the exclusive rights of reproduction, storage, communication to the public, and making available of the works, even in partial or fragmented form. In the plaintiffs' view, any use of the content - including acquisition, storage, and processing for algorithmic training purposes - requires the prior consent of the rights holder.

On this point, the plaintiffs clarify that the training of generative AI models cannot be considered a merely technical or internal use but rather constitutes a genuine form of economic exploitation of the works. The model, in fact, assimilates and internalizes narrative structures, expressive elements, and audiovisual sequences, which are then reused to generate outputs intended for the public and integrated into a commercialized service. This mechanism establishes a direct causal nexus between the use of the protected works and the economic value of the service offered by Perplexity, constituting a legally relevant use under Articles 13 and 16 LDA.

The technical verifications conducted by the plaintiffs have demonstrated – with direct evidence and textual admissions from the Perplexity AI platform itself – that Perplexity has massively acquired, copied and stored works to use them as training materials for its generative AI, thus employing them as input to produce derivative content and to run a commercial service, which is also partially fees-based. The unauthorized use of these works for training purposes and for generating outputs that incorporate protected elements would amount to a violation of the plaintiffs’ exclusive rights and an undue appropriation of their creative and economic investments.

In light of this, again according to the plaintiffs' allegations, Perplexity cannot invoke the text and data mining exception provided for in Article 4 of the DSM Directive and Article 70-quater of the LDA. The defendant's conduct, as stated in the complaint, is incompatible with the principles enshrined in the Berne Convention, particularly with the three-step test of Article 9, which allows for copyright exceptions only in special cases, provided that the normal exploitation of the work is not compromised and no unreasonable prejudice is caused to the rights holders. According to the plaintiffs, Perplexity's use is not a limited use in a special case, but rather a massive and generalized use; it directly affects the normal exploitation of the works by subtracting economic value from the original content; and it causes unjustified prejudice to the rights holders, who see their works used to power competing services without any form of license or compensation. The perspective of the supported legal-logical framework, therefore, is in full continuity with the European case law that imposes a restrictive interpretation of copyright exceptions.

The damage arises in re ipsa from the infringement of exclusive rights and manifests itself in the loss of control over the exploitation of the works, the compromise of the economic value of the content, and the diversion of traffic and audience to a third-party service that feeds on the creative work of others.

In this context, the plaintiffs are asking the Court of Rome to ascertain the unlawfulness of the conduct and to issue appropriate injunctions, together with the publication of the decision and an award of damages, to be determined based on the criteria of the price of consent and of the disgorgement of profits.

Image by Kohji Asakawa from Pixabay